- 學位論文
多來源之網路輿論關鍵字分析之研究
Research on Keyword Analysis of Internet Public Opinion from Multiple Sources
摘要
網路作為各種資訊的匯流，也包含了各式議題的討論和走向，在網路世界延燒的議題，往往最直接能看出大部分民眾的想法。藉由社群媒體，我們可以更容易發表看法，但也更輕易地受到社群媒體上人們的話語和行為所影響。本研究主要目的分兩個方向，一是判斷輿情，制定相關市場、政策規劃，二是事件發生後，整理事件輿論，選擇符合大眾期待的危機處理。目前有許多商品或企業會製作問卷了解市場接受度，但問卷的內容會受到設計者的影響，設計問卷者是否完全熟悉、回收的樣本數、收集的時間、受試者是否對問卷誠實且耐心的填寫完整問卷，都是可能的變因，綜合以上因素市場問卷調查的問題。在網路世界中，大多人能暢所欲言表達真實的看法，是最能貼近大眾看法的方式。因此由網路輿論獲取大眾事件的看法，幫助有關單位作出事件的危機處理，或對政策實施方向調整，甚至更進一步利用從眾效應主導事件。 透過編寫自動爬蟲從網路上收集各式新聞及社群媒體上的討論和報導，以TextRank演算法找出各文章中關鍵字，將所有文章關鍵字統計以觀察各平台熱門主題關鍵字，並以簡化的方法更快的找出關鍵字趨勢，最後對比入口網站關鍵字查詢量評估兩者之間的差異，以更多方面的監測的輿論趨勢發生時間點。
並列摘要
As a confluence of all kinds of information, the Internet also includes discussions and trends of various issues. Issues that are spreading in the online world can often be the most direct way to see the thoughts of most people. Through social media, we can more easily express our opinions, but also more easily be influenced by what people say and do on social media. The main purpose of this research is divided into two directions. One is to judge public opinion and formulate relevant market and policy plans. The other is to sort out the public opinion of the incident after the incident and choose the crisis management that meets the public's expectations. At present, many products or enterprises will make questionnaires to understand the market acceptance, but the content of the questionnaire will be affected by the designer, whether the designer of the questionnaire is completely familiar, the number of samples recovered, the time of collection, and whether the subjects are honest and patient with the questionnaire. The completed questionnaires are all possible variables, and the above factors are the questions of the market questionnaire survey. In the online world, most people can express their true opinions freely, which is the most close to the public opinion. Therefore, online public opinion can obtain opinions on popular events, help relevant units to deal with the crisis of the event, or adjust the direction of policy implementation, and even further use the herd effect to dominate the event. By writing automatic crawlers to collect various news and social media discussions and reports from the Internet, use TextRank algorithm to find keywords in each article, count all article keywords to observe popular topic keywords on each platform, and use TextRank algorithm to find keywords in each article. The simplified method finds out the keyword trend faster, and finally compares the keyword query volume of the portal website to evaluate the difference between the two, so as to monitor the occurrence time point of the public opinion trend in more aspects.